import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns

from DataProcessingMethods.DataPrepMultiClassv1 import prepare_data_pipeline

# Load the Forest Cover Type dataset
dataset_name = "Forest Cover Type Dataset"
df = pd.read_csv("Datasets/covtype.csv", na_values="?")  # Adjust path if needed

# Display basic information about the dataset
print(f"\n{dataset_name} shape: {df.shape}")
print(f"Number of classes in Cover_Type: {df['Cover_Type'].nunique()}")
print(f"Class distribution:\n{df['Cover_Type'].value_counts()}")
print("\nSample data:")
print(df.head())

# Define features for analysis
# Specify only the continuous numeric features for outlier detection
numeric_features = [
    'Elevation', 'Aspect', 'Slope', 
    'Horizontal_Distance_To_Hydrology', 'Vertical_Distance_To_Hydrology', 
    'Horizontal_Distance_To_Roadways', 
    'Hillshade_9am', 'Hillshade_Noon', 'Hillshade_3pm', 
    'Horizontal_Distance_To_Fire_Points'
]

# Features to display distributions for
dist_features = [
    'Slope', 'Horizontal_Distance_To_Hydrology', 'Vertical_Distance_To_Hydrology', 
    'Horizontal_Distance_To_Roadways', 'Hillshade_9am', 'Hillshade_Noon', 'Hillshade_3pm', 
    'Horizontal_Distance_To_Fire_Points'
]

# Target column
target = "Cover_Type"

# Since you don't want outlier removal for binary features (0/1),
# only include continuous features in the outlier detection
outlier_features = [
    'Elevation', 'Aspect', 'Slope', 
    'Horizontal_Distance_To_Hydrology', 'Vertical_Distance_To_Hydrology', 
    'Horizontal_Distance_To_Roadways', 
    'Hillshade_9am', 'Hillshade_Noon', 'Hillshade_3pm', 
    'Horizontal_Distance_To_Fire_Points'
]

# Run the data preparation pipeline
print("\nRunning data preparation pipeline...")
#df_no_outliers = prepare_data_pipeline(
#    df=df,
#    list_features_specialise_outliers=outlier_features,  # Only continuous features for outlier detection
#    numeric_features=numeric_features,
#    dist_features=dist_features,
#    target=target
#)

# Print summary of results
#print("\nPrepared dataset shape:", df_no_outliers.shape)
#print("\nClass distribution after preparation:")
#print(df_no_outliers[target].value_counts())

print("\nDone! Check for generated visualizations.")

Forest Cover Type Dataset shape: (581012, 55)
Number of classes in Cover_Type: 7
Class distribution:
Cover_Type
2    283301
1    211840
3     35754
7     20510
6     17367
5      9493
4      2747
Name: count, dtype: int64

Sample data:
   Elevation  Aspect  Slope  Horizontal_Distance_To_Hydrology  \
0       2596      51      3                               258   
1       2590      56      2                               212   
2       2804     139      9                               268   
3       2785     155     18                               242   
4       2595      45      2                               153   

   Vertical_Distance_To_Hydrology  Horizontal_Distance_To_Roadways  \
0                               0                              510   
1                              -6                              390   
2                              65                             3180   
3                             118                             3090   
4                              -1                              391   

   Hillshade_9am  Hillshade_Noon  Hillshade_3pm  \
0            221             232            148   
1            220             235            151   
2            234             238            135   
3            238             238            122   
4            220             234            150   

   Horizontal_Distance_To_Fire_Points  ...  Soil_Type32  Soil_Type33  \
0                                6279  ...            0            0   
1                                6225  ...            0            0   
2                                6121  ...            0            0   
3                                6211  ...            0            0   
4                                6172  ...            0            0   

   Soil_Type34  Soil_Type35  Soil_Type36  Soil_Type37  Soil_Type38  \
0            0            0            0            0            0   
1            0            0            0            0            0   
2            0            0            0            0            0   
3            0            0            0            0            0   
4            0            0            0            0            0   

   Soil_Type39  Soil_Type40  Cover_Type  
0            0            0           5  
1            0            0           5  
2            0            0           2  
3            0            0           2  
4            0            0           5  

[5 rows x 55 columns]

Running data preparation pipeline...

Done! Check for generated visualizations.

df_no_outliers.to_csv("Datasets/covtypeCLEANED.csv", index=False)
print("Done")

Done

from GenerationMethods.MultiClassification.MultiVAE2 import augment_dataframe_vae_enhanced
df_no_outliers = pd.read_csv("Datasets/covtypeCLEANED.csv")
# Run the augmentation
original_train, augmented_train, test_set, success = augment_dataframe_vae_enhanced(
    df=df_no_outliers,
    target='Cover_Type',
    test_size=0.25,
    random_state=42, 
    n_classes_to_augment=4, 
    ratio_limit=0.5,
    diminishing_factor=0.65,
    vae_epochs=2,             
    vae_batch_size=64,         
    latent_dim=48,
    hidden_dims=[512, 256, 128],
    temperature=0.8,
    matching_factor=0.25,
    early_stopping_patience=50  
)


if success:
    # CORRECT USAGE - This gets numeric features directly from the returned DataFrame
    numeric_features = augmented_train.select_dtypes(include=['number']).columns
    
    # Exclude the target and synthetic columns if they exist and are numeric
    columns_to_exclude = []
    if 'quality' in numeric_features:
        columns_to_exclude.append('quality')
    if 'synthetic' in numeric_features:
        columns_to_exclude.append('synthetic')
        
    # Filter numeric features to exclude certain columns
    if columns_to_exclude:
        numeric_features = [col for col in numeric_features if col not in columns_to_exclude]
    
    # Round numeric features
    
    augmented_train[numeric_features] = augmented_train[numeric_features].round(2)
    
    # Save outputs
    original_train.to_csv("OutputTrainingSets/original_trainVAEForestFINAL.csv", index=False)
    augmented_train.to_csv("OutputTrainingSets/augmented_trainVAEForestFINAL.csv", index=False)
    test_set.to_csv("OutputTrainingSets/test_setVAEForestFINAL.csv", index=False)
else:
    print("Augmentation failed. Check the error messages.")

2025-04-14 14:39:43.462946: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:479] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered
2025-04-14 14:39:43.489860: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:10575] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
2025-04-14 14:39:43.489913: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1442] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
2025-04-14 14:39:43.507414: I tensorflow/core/platform/cpu_feature_guard.cc:210] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 AVX512F FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
2025-04-14 14:39:44.597447: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT

Starting enhanced VAE-based augmentation...
Target values range: [1, 7], unique classes: 7
Warning: max_label 7 >= num_classes 7, adjusting num_classes to 8
Training data shape: (427285, 54)
Test data shape: (142429, 54)
Training enhanced VAE model...
Using 8 classes for one-hot encoding (max label: 7)

2025-04-14 14:39:48.398834: E external/local_xla/xla/stream_executor/cuda/cuda_driver.cc:282] failed call to cuInit: CUDA_ERROR_NO_DEVICE: no CUDA-capable device is detected

Epoch 1/2
5675/5675 ━━━━━━━━━━━━━━━━━━━━ 103s 17ms/step - loss: 0.2908 - val_loss: 0.1185
Epoch 2/2
5675/5675 ━━━━━━━━━━━━━━━━━━━━ 95s 17ms/step - loss: 0.1187 - val_loss: 0.1100
Restoring model weights from the end of the best epoch: 2.
Could not load best model weights, using final weights
Class distribution: {1: 155577, 2: 208214, 3: 26803, 4: 2060, 5: 6808, 6: 13003, 7: 14820}
Classes to augment (from smallest to largest): [4, 5, 6, 7]
Class 4: Generating 1030 synthetic samples
  - Current count: 2060
  - Diminishing factor: 1.00
  - Theoretical target count: 208214
  - Limited by ratio constraint: max 1030 new samples
  - Final target count: 3090
65/65 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step
33/33 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step
Class 5: Generating 3404 synthetic samples
  - Current count: 6808
  - Diminishing factor: 0.65
  - Theoretical target count: 135339
  - Limited by ratio constraint: max 3404 new samples
  - Final target count: 10212
213/213 ━━━━━━━━━━━━━━━━━━━━ 1s 3ms/step
107/107 ━━━━━━━━━━━━━━━━━━━━ 0s 2ms/step
Class 6: Generating 6501 synthetic samples
  - Current count: 13003
  - Diminishing factor: 0.42
  - Theoretical target count: 87970
  - Limited by ratio constraint: max 6501 new samples
  - Final target count: 19504
407/407 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
204/204 ━━━━━━━━━━━━━━━━━━━━ 0s 2ms/step
Class 7: Generating 7410 synthetic samples
  - Current count: 14820
  - Diminishing factor: 0.27
  - Theoretical target count: 57180
  - Limited by ratio constraint: max 7410 new samples
  - Final target count: 22230
464/464 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
232/232 ━━━━━━━━━━━━━━━━━━━━ 1s 2ms/step
Debug - Synthetic samples shape before inverse transform: (18345, 54)
Debug - Synthetic samples range: [2.799956519083935e-06, 0.9997535943984985]
Applying quantile matching to synthetic samples...
Original DataFrame info:
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 427285 entries, 0 to 427284
Data columns (total 56 columns):
 #   Column                              Non-Null Count   Dtype  
---  ------                              --------------   -----  
 0   Elevation                           427285 non-null  float64
 1   Aspect                              427285 non-null  float64
 2   Slope                               427285 non-null  float64
 3   Horizontal_Distance_To_Hydrology    427285 non-null  float64
 4   Vertical_Distance_To_Hydrology      427285 non-null  float64
 5   Horizontal_Distance_To_Roadways     427285 non-null  float64
 6   Hillshade_9am                       427285 non-null  float64
 7   Hillshade_Noon                      427285 non-null  float64
 8   Hillshade_3pm                       427285 non-null  float64
 9   Horizontal_Distance_To_Fire_Points  427285 non-null  float64
 10  Wilderness_Area1                    427285 non-null  float64
 11  Wilderness_Area2                    427285 non-null  float64
 12  Wilderness_Area3                    427285 non-null  float64
 13  Wilderness_Area4                    427285 non-null  float64
 14  Soil_Type1                          427285 non-null  float64
 15  Soil_Type2                          427285 non-null  float64
 16  Soil_Type3                          427285 non-null  float64
 17  Soil_Type4                          427285 non-null  float64
 18  Soil_Type5                          427285 non-null  float64
 19  Soil_Type6                          427285 non-null  float64
 20  Soil_Type7                          427285 non-null  float64
 21  Soil_Type8                          427285 non-null  float64
 22  Soil_Type9                          427285 non-null  float64
 23  Soil_Type10                         427285 non-null  float64
 24  Soil_Type11                         427285 non-null  float64
 25  Soil_Type12                         427285 non-null  float64
 26  Soil_Type13                         427285 non-null  float64
 27  Soil_Type14                         427285 non-null  float64
 28  Soil_Type15                         427285 non-null  float64
 29  Soil_Type16                         427285 non-null  float64
 30  Soil_Type17                         427285 non-null  float64
 31  Soil_Type18                         427285 non-null  float64
 32  Soil_Type19                         427285 non-null  float64
 33  Soil_Type20                         427285 non-null  float64
 34  Soil_Type21                         427285 non-null  float64
 35  Soil_Type22                         427285 non-null  float64
 36  Soil_Type23                         427285 non-null  float64
 37  Soil_Type24                         427285 non-null  float64
 38  Soil_Type25                         427285 non-null  float64
 39  Soil_Type26                         427285 non-null  float64
 40  Soil_Type27                         427285 non-null  float64
 41  Soil_Type28                         427285 non-null  float64
 42  Soil_Type29                         427285 non-null  float64
 43  Soil_Type30                         427285 non-null  float64
 44  Soil_Type31                         427285 non-null  float64
 45  Soil_Type32                         427285 non-null  float64
 46  Soil_Type33                         427285 non-null  float64
 47  Soil_Type34                         427285 non-null  float64
 48  Soil_Type35                         427285 non-null  float64
 49  Soil_Type36                         427285 non-null  float64
 50  Soil_Type37                         427285 non-null  float64
 51  Soil_Type38                         427285 non-null  float64
 52  Soil_Type39                         427285 non-null  float64
 53  Soil_Type40                         427285 non-null  float64
 54  Cover_Type                          427285 non-null  int64  
 55  synthetic                           427285 non-null  bool   
dtypes: bool(1), float64(54), int64(1)
memory usage: 179.7 MB
None

Synthetic DataFrame info:
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 18345 entries, 0 to 18344
Data columns (total 56 columns):
 #   Column                              Non-Null Count  Dtype  
---  ------                              --------------  -----  
 0   Elevation                           18345 non-null  float64
 1   Aspect                              18345 non-null  float64
 2   Slope                               18345 non-null  float64
 3   Horizontal_Distance_To_Hydrology    18345 non-null  float64
 4   Vertical_Distance_To_Hydrology      18345 non-null  float64
 5   Horizontal_Distance_To_Roadways     18345 non-null  float64
 6   Hillshade_9am                       18345 non-null  float64
 7   Hillshade_Noon                      18345 non-null  float64
 8   Hillshade_3pm                       18345 non-null  float64
 9   Horizontal_Distance_To_Fire_Points  18345 non-null  float64
 10  Wilderness_Area1                    18345 non-null  float64
 11  Wilderness_Area2                    18345 non-null  float64
 12  Wilderness_Area3                    18345 non-null  float64
 13  Wilderness_Area4                    18345 non-null  float64
 14  Soil_Type1                          18345 non-null  float64
 15  Soil_Type2                          18345 non-null  float64
 16  Soil_Type3                          18345 non-null  float64
 17  Soil_Type4                          18345 non-null  float64
 18  Soil_Type5                          18345 non-null  float64
 19  Soil_Type6                          18345 non-null  float64
 20  Soil_Type7                          18345 non-null  float64
 21  Soil_Type8                          18345 non-null  float64
 22  Soil_Type9                          18345 non-null  float64
 23  Soil_Type10                         18345 non-null  float64
 24  Soil_Type11                         18345 non-null  float64
 25  Soil_Type12                         18345 non-null  float64
 26  Soil_Type13                         18345 non-null  float64
 27  Soil_Type14                         18345 non-null  float64
 28  Soil_Type15                         18345 non-null  float64
 29  Soil_Type16                         18345 non-null  float64
 30  Soil_Type17                         18345 non-null  float64
 31  Soil_Type18                         18345 non-null  float64
 32  Soil_Type19                         18345 non-null  float64
 33  Soil_Type20                         18345 non-null  float64
 34  Soil_Type21                         18345 non-null  float64
 35  Soil_Type22                         18345 non-null  float64
 36  Soil_Type23                         18345 non-null  float64
 37  Soil_Type24                         18345 non-null  float64
 38  Soil_Type25                         18345 non-null  float64
 39  Soil_Type26                         18345 non-null  float64
 40  Soil_Type27                         18345 non-null  float64
 41  Soil_Type28                         18345 non-null  float64
 42  Soil_Type29                         18345 non-null  float64
 43  Soil_Type30                         18345 non-null  float64
 44  Soil_Type31                         18345 non-null  float64
 45  Soil_Type32                         18345 non-null  float64
 46  Soil_Type33                         18345 non-null  float64
 47  Soil_Type34                         18345 non-null  float64
 48  Soil_Type35                         18345 non-null  float64
 49  Soil_Type36                         18345 non-null  float64
 50  Soil_Type37                         18345 non-null  float64
 51  Soil_Type38                         18345 non-null  float64
 52  Soil_Type39                         18345 non-null  float64
 53  Soil_Type40                         18345 non-null  float64
 54  Cover_Type                          18345 non-null  int64  
 55  synthetic                           18345 non-null  bool   
dtypes: bool(1), float64(54), int64(1)
memory usage: 7.7 MB
None

Augmented DataFrame info:
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 445630 entries, 0 to 445629
Data columns (total 56 columns):
 #   Column                              Non-Null Count   Dtype  
---  ------                              --------------   -----  
 0   Elevation                           445630 non-null  float64
 1   Aspect                              445630 non-null  float64
 2   Slope                               445630 non-null  float64
 3   Horizontal_Distance_To_Hydrology    445630 non-null  float64
 4   Vertical_Distance_To_Hydrology      445630 non-null  float64
 5   Horizontal_Distance_To_Roadways     445630 non-null  float64
 6   Hillshade_9am                       445630 non-null  float64
 7   Hillshade_Noon                      445630 non-null  float64
 8   Hillshade_3pm                       445630 non-null  float64
 9   Horizontal_Distance_To_Fire_Points  445630 non-null  float64
 10  Wilderness_Area1                    445630 non-null  float64
 11  Wilderness_Area2                    445630 non-null  float64
 12  Wilderness_Area3                    445630 non-null  float64
 13  Wilderness_Area4                    445630 non-null  float64
 14  Soil_Type1                          445630 non-null  float64
 15  Soil_Type2                          445630 non-null  float64
 16  Soil_Type3                          445630 non-null  float64
 17  Soil_Type4                          445630 non-null  float64
 18  Soil_Type5                          445630 non-null  float64
 19  Soil_Type6                          445630 non-null  float64
 20  Soil_Type7                          445630 non-null  float64
 21  Soil_Type8                          445630 non-null  float64
 22  Soil_Type9                          445630 non-null  float64
 23  Soil_Type10                         445630 non-null  float64
 24  Soil_Type11                         445630 non-null  float64
 25  Soil_Type12                         445630 non-null  float64
 26  Soil_Type13                         445630 non-null  float64
 27  Soil_Type14                         445630 non-null  float64
 28  Soil_Type15                         445630 non-null  float64
 29  Soil_Type16                         445630 non-null  float64
 30  Soil_Type17                         445630 non-null  float64
 31  Soil_Type18                         445630 non-null  float64
 32  Soil_Type19                         445630 non-null  float64
 33  Soil_Type20                         445630 non-null  float64
 34  Soil_Type21                         445630 non-null  float64
 35  Soil_Type22                         445630 non-null  float64
 36  Soil_Type23                         445630 non-null  float64
 37  Soil_Type24                         445630 non-null  float64
 38  Soil_Type25                         445630 non-null  float64
 39  Soil_Type26                         445630 non-null  float64
 40  Soil_Type27                         445630 non-null  float64
 41  Soil_Type28                         445630 non-null  float64
 42  Soil_Type29                         445630 non-null  float64
 43  Soil_Type30                         445630 non-null  float64
 44  Soil_Type31                         445630 non-null  float64
 45  Soil_Type32                         445630 non-null  float64
 46  Soil_Type33                         445630 non-null  float64
 47  Soil_Type34                         445630 non-null  float64
 48  Soil_Type35                         445630 non-null  float64
 49  Soil_Type36                         445630 non-null  float64
 50  Soil_Type37                         445630 non-null  float64
 51  Soil_Type38                         445630 non-null  float64
 52  Soil_Type39                         445630 non-null  float64
 53  Soil_Type40                         445630 non-null  float64
 54  Cover_Type                          445630 non-null  int64  
 55  synthetic                           445630 non-null  bool   
dtypes: bool(1), float64(54), int64(1)
memory usage: 187.4 MB
None
Augmented data summary:
  - Original samples: 427285
  - Synthetic samples: 18345
  - Total samples: 445630
Class distribution before: {1: 155577, 2: 208214, 3: 26803, 4: 2060, 5: 6808, 6: 13003, 7: 14820}
Class distribution after: {1: 155577, 2: 208214, 3: 26803, 4: 3090, 5: 10212, 6: 19504, 7: 22230}

Saved results to ./output/
1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 189ms/step

<Figure size 1000x600 with 0 Axes>

from ValidationMethods.MultiClassValidation import validate_synthetic_data_per_class, analyze_target_distribution
from GenerationMethods.MultiClassification.undersampleMajorityClasses import undersample_majority_classes
import pandas as pd

# Load the CSV files generated by the augmentation process.
# (These files are assumed to have been generated using an oversampling method adapted for multi-class data.)
original_train2 = pd.read_csv("OutputTrainingSets/original_trainVAEForestFINAL.csv")
augmented_train2 = pd.read_csv("OutputTrainingSets/augmented_trainVAEForestFINAL.csv")
test_set2 = pd.read_csv("OutputTrainingSets/test_setVAEForestFINAL.csv")


# Apply undersampling before augmentation
df_augmented_train2 = undersample_majority_classes(augmented_train2, 
                                              majority_classes=[1, 2], 
                                              target_ratio=0.15)
augmented_train2.to_csv("OutputTrainingSets/augmented_trainVAEForestFINAL.csv", index=False)

#print(flag)
#augmented_train2[numeric_features] = augmented_train2[numeric_features].round(2)

# Define the columns to keep: continuous features + target.
continuous_features = ['Elevation', 'Aspect', 'Slope', 'Horizontal_Distance_To_Hydrology', 'Vertical_Distance_To_Hydrology', 'Horizontal_Distance_To_Roadways', 'Hillshade_9am', 'Hillshade_Noon', 'Hillshade_3pm', 'Horizontal_Distance_To_Fire_Points', 'Wilderness_Area1', 'Wilderness_Area2', 'Wilderness_Area3', 'Wilderness_Area4', 'Soil_Type1', 'Soil_Type2', 'Soil_Type3', 'Soil_Type4', 'Soil_Type5', 'Soil_Type6', 'Soil_Type7', 'Soil_Type8', 'Soil_Type9', 'Soil_Type10', 'Soil_Type11', 'Soil_Type12', 'Soil_Type13', 'Soil_Type14', 'Soil_Type15', 'Soil_Type16', 'Soil_Type17', 'Soil_Type18', 'Soil_Type19', 'Soil_Type20', 'Soil_Type21', 'Soil_Type22', 'Soil_Type23', 'Soil_Type24', 'Soil_Type25', 'Soil_Type26', 'Soil_Type27', 'Soil_Type28', 'Soil_Type29', 'Soil_Type30', 'Soil_Type31', 'Soil_Type32', 'Soil_Type33', 'Soil_Type34', 'Soil_Type35', 'Soil_Type36', 'Soil_Type37', 'Soil_Type38', 'Soil_Type39', 'Soil_Type40']
categorical_features = ["Cover_Type"]
cols_to_keep = continuous_features + categorical_features

# Keep only the desired columns and drop rows with missing values.
original_train2 = original_train2[cols_to_keep].dropna()
# For the augmented training set, also keep the "synthetic" column.
augmented_train2 = augmented_train2[cols_to_keep + ['synthetic']].dropna()
test_set2 = test_set2[cols_to_keep].dropna()
    
target = "Cover_Type"

# Extract synthetic samples from the augmented training set (synthetic == True).
synthetic_samples = augmented_train2[augmented_train2['synthetic'] == True]

# Run the validation analysis comparing original training data against synthetic samples.
metrics = validate_synthetic_data_per_class(
    original=original_train2,
    synthetic=synthetic_samples,
    continuous_features=continuous_features,
    categorical_features=categorical_features,
    target = target,
    num_classes = 4,
    distance_threshold=0.5,
    density_threshold=0.5,
    gamma=1.0,
    plot=True
)
    
print("Validation metrics:")
print(metrics)
    
print("\n### Target Distribution Analysis on Original Training Set ###")
analyze_target_distribution(original_train2, target=target)
print("\n### Target Distribution Analysis on Augmented Training Set ###")
analyze_target_distribution(augmented_train2, target=target)
print("\n### Target Distribution Analysis on Test Set ###")
analyze_target_distribution(test_set2, target=target)


### Validation for class: 4 ###

### Continuous Features Validation ###

Feature: Elevation
Original: mean=2224.681, std=101.540
Synthetic: mean=2552.035, std=244.547
KS test: statistic=0.700, p-value=0.000
-> Significant difference detected!
Synthetic shape after dropna: (1030, 56)
Any NA in mcg for synthetic? 0
Synthetic mcg describe:
 count    1030.000000
mean     2552.034553
std       244.546754
min      1869.240000
25%      2372.165000
50%      2587.895000
75%      2747.302500
max      3021.000000
Name: Elevation, dtype: float64

Feature: Aspect
Original: mean=137.600, std=87.461
Synthetic: mean=115.970, std=88.486
KS test: statistic=0.211, p-value=0.000
-> Significant difference detected!
Synthetic shape after dropna: (1030, 56)
Any NA in mcg for synthetic? 0
Synthetic mcg describe:
 count    1030.000000
mean      115.969981
std        88.485571
min         0.000000
25%        45.610000
50%        94.175000
75%       165.000000
max       358.000000
Name: Aspect, dtype: float64

Feature: Slope
Original: mean=18.610, std=9.280
Synthetic: mean=17.368, std=6.877
KS test: statistic=0.140, p-value=0.000
-> Significant difference detected!
Synthetic shape after dropna: (1030, 56)
Any NA in mcg for synthetic? 0
Synthetic mcg describe:
 count    1030.000000
mean       17.367835
std         6.876850
min         2.000000
25%        12.000000
50%        16.305000
75%        22.000000
max        50.000000
Name: Slope, dtype: float64

Feature: Horizontal_Distance_To_Hydrology
Original: mean=105.259, std=139.052
Synthetic: mean=94.667, std=94.635
KS test: statistic=0.231, p-value=0.000
-> Significant difference detected!
Synthetic shape after dropna: (1030, 56)
Any NA in mcg for synthetic? 0
Synthetic mcg describe:
 count    1030.000000
mean       94.667136
std        94.634996
min         0.000000
25%        30.000000
50%        67.000000
75%       150.000000
max       621.890000
Name: Horizontal_Distance_To_Hydrology, dtype: float64

Feature: Vertical_Distance_To_Hydrology
Original: mean=40.335, std=58.823
Synthetic: mean=22.769, std=36.872
KS test: statistic=0.202, p-value=0.000
-> Significant difference detected!
Synthetic shape after dropna: (1030, 56)
Any NA in mcg for synthetic? 0
Synthetic mcg describe:
 count    1030.000000
mean       22.769379
std        36.872228
min       -87.450000
25%         0.262500
50%        13.000000
75%        37.000000
max       242.100000
Name: Vertical_Distance_To_Hydrology, dtype: float64

Feature: Horizontal_Distance_To_Roadways
Original: mean=920.270, std=368.143
Synthetic: mean=1128.825, std=817.765
KS test: statistic=0.226, p-value=0.000
-> Significant difference detected!
Synthetic shape after dropna: (1030, 56)
Any NA in mcg for synthetic? 0
Synthetic mcg describe:
 count    1030.000000
mean     1128.825136
std       817.765398
min        25.160000
25%       551.455000
50%       962.455000
75%      1483.800000
max      6056.680000
Name: Horizontal_Distance_To_Roadways, dtype: float64

Feature: Hillshade_9am
Original: mean=228.063, std=24.382
Synthetic: mean=224.608, std=17.487
KS test: statistic=0.244, p-value=0.000
-> Significant difference detected!
Synthetic shape after dropna: (1030, 56)
Any NA in mcg for synthetic? 0
Synthetic mcg describe:
 count    1030.000000
mean      224.607864
std        17.487470
min       157.000000
25%       215.000000
50%       227.000000
75%       237.000000
max       254.000000
Name: Hillshade_9am, dtype: float64

Feature: Hillshade_Noon
Original: mean=216.899, std=20.883
Synthetic: mean=220.407, std=19.620
KS test: statistic=0.093, p-value=0.000
-> Significant difference detected!
Synthetic shape after dropna: (1030, 56)
Any NA in mcg for synthetic? 0
Synthetic mcg describe:
 count    1030.000000
mean      220.407398
std        19.620388
min        98.450000
25%       210.000000
50%       223.000000
75%       234.000000
max       254.000000
Name: Hillshade_Noon, dtype: float64

Feature: Hillshade_3pm
Original: mean=111.642, std=49.051
Synthetic: mean=115.813, std=35.778
KS test: statistic=0.147, p-value=0.000
-> Significant difference detected!
Synthetic shape after dropna: (1030, 56)
Any NA in mcg for synthetic? 0
Synthetic mcg describe:
 count    1030.00000
mean      115.81266
std        35.77751
min         0.00000
25%        96.00000
50%       117.70500
75%       139.00000
max       213.04000
Name: Hillshade_3pm, dtype: float64

Feature: Horizontal_Distance_To_Fire_Points
Original: mean=860.698, std=480.851
Synthetic: mean=1193.648, std=773.064
KS test: statistic=0.172, p-value=0.000
-> Significant difference detected!
Synthetic shape after dropna: (1030, 56)
Any NA in mcg for synthetic? 0
Synthetic mcg describe:
 count    1030.000000
mean     1193.648291
std       773.064062
min        55.550000
25%       633.000000
50%      1036.760000
75%      1599.892500
max      5673.380000
Name: Horizontal_Distance_To_Fire_Points, dtype: float64

import pandas as pd
from MachineLearningModels.MultiClassLogisticRegression import train_logistic_model, evaluate_model, compare_models
from sklearn.preprocessing import LabelEncoder
import numpy as np

# Load CSV files for the Yeast dataset.
original_train = pd.read_csv("OutputTrainingSets/original_trainVAEForestFINAL.csv")
augmented_train = pd.read_csv("OutputTrainingSets/augmented_trainVAEForestFINAL.csv")
test_set = pd.read_csv("OutputTrainingSets/test_setVAEForestFINAL.csv")

# Define the feature columns and target.
features = numeric_features
#          OR
#features = continuous_features

categorical_features = ["Cover_Type"]

# Re-map the target column so that labels are contiguous (0, 1, 2, ...).
le = LabelEncoder()
original_train[target] = le.fit_transform(original_train[target])
augmented_train[target] = le.transform(augmented_train[target])
test_set[target] = le.transform(test_set[target])

# Train a Logistic Regression model on the original training dataset.
model_original = train_logistic_model(original_train, features, target)
metrics_original = evaluate_model(model_original, test_set, features, target)

# Train a Logistic Regression model on the augmented training dataset.
model_augmented = train_logistic_model(augmented_train, features, target)
metrics_augmented = evaluate_model(model_augmented, test_set, features, target)

unique_labels = np.sort(metrics_original['y_test'].unique())
# Now, inverse transform these labels using the same LabelEncoder 'le' used earlier
target_names = [str(x) for x in le.inverse_transform(unique_labels)]

# Compare the performance of the two models.
compare_models(metrics_original, metrics_augmented, target_names)

=== Original Training Model Metrics ===
Accuracy: 0.720
AUC: 0.930
Classification Report:
              precision    recall  f1-score   support

           1       0.72      0.74      0.73     51859
           2       0.73      0.80      0.76     69405
           3       0.61      0.65      0.63      8935
           4       0.43      0.34      0.38       686
           5       0.67      0.02      0.04      2269
           6       0.42      0.12      0.18      4335
           7       0.97      0.37      0.54      4940

    accuracy                           0.72    142429
   macro avg       0.65      0.43      0.47    142429
weighted avg       0.72      0.72      0.71    142429

Confusion Matrix:
[[38574 13257     8     0     0     0    20]
 [12426 55499  1336     0    21   106    17]
 [    1  2263  5806   258     0   607     0]
 [    0    20   433   231     0     2     0]
 [    1  2159    54     0    42     0    13]
 [    1  1841  1929    48     0   514     2]
 [ 2376   730     0     0     0     0  1834]]

=== Augmented Training Model Metrics ===
Accuracy: 0.736
AUC: 0.937
Classification Report:
              precision    recall  f1-score   support

           1       0.74      0.74      0.74     51859
           2       0.74      0.78      0.76     69405
           3       0.64      0.60      0.62      8935
           4       0.38      0.52      0.44       686
           5       1.00      0.00      0.00      2269
           6       0.52      0.39      0.45      4335
           7       1.00      0.92      0.96      4940

    accuracy                           0.74    142429
   macro avg       0.72      0.56      0.57    142429
weighted avg       0.74      0.74      0.73    142429

Confusion Matrix:
[[38380 13457     9    12     0     0     1]
 [13401 54442  1283    13     0   253    13]
 [    1  1890  5344   458     0  1242     0]
 [    0     5   294   356     0    31     0]
 [    6  2190    61     0     2     9     1]
 [    0  1143  1396   106     0  1687     3]
 [   63   330     0     0     0     0  4547]]

"""
import pandas as pd
from MachineLearningModels.MultiClassXGBoostComparison import train_xgb_model, evaluate_model, compare_models
from sklearn.preprocessing import LabelEncoder
import numpy as np

# Load CSV files for the Yeast dataset.
original_train = pd.read_csv("OutputTrainingSets/original_trainVAEForest2.csv")
augmented_train = pd.read_csv("OutputTrainingSets/augmented_trainVAEForest2.csv")
test_set = pd.read_csv("OutputTrainingSets/test_setVAEForest2.csv")
    
# Define the feature columns and target.
# For the Yeast dataset, typical numeric features are:
features = numeric_features
categorical_features = ["Cover_Type"]

# Re-map the target column so that labels are contiguous integers (0, 1, 2, …)
le = LabelEncoder()
original_train[target] = le.fit_transform(original_train[target])
augmented_train[target] = le.transform(augmented_train[target])
test_set[target] = le.transform(test_set[target])
    
# Train a multi-class XGBoost model on the original training dataset.
model_original = train_xgb_model(original_train, features, target)
metrics_original = evaluate_model(model_original, test_set, features, target)
    
# Train a multi-class XGBoost model on the augmented training dataset.
model_augmented = train_xgb_model(augmented_train, features, target)
metrics_augmented = evaluate_model(model_augmented, test_set, features, target)

unique_labels = np.sort(metrics_original['y_test'].unique())
# Now, inverse transform these labels using the same LabelEncoder 'le' used earlier
target_names = [str(x) for x in le.inverse_transform(unique_labels)]

# Compare the performance of the two models.
compare_models(metrics_original, metrics_augmented, target_names)
"""

"""
import pandas as pd
from MachineLearningModels.MultiClassRandomForest import train_rf_model, evaluate_model, compare_models
from sklearn.preprocessing import LabelEncoder
import numpy as np

# Load CSV files for the Yeast dataset.
original_train = pd.read_csv("OutputTrainingSets/original_trainVAEForest.csv")
augmented_train = pd.read_csv("OutputTrainingSets/augmented_trainVAEForest.csv")
test_set = pd.read_csv("OutputTrainingSets/test_setVAEForest.csv")

# Define the feature columns and target.
features = numeric_features
categorical_features = ["Cover_Type"]

# Re-map the target column so that labels are contiguous.
le = LabelEncoder()
original_train[target] = le.fit_transform(original_train[target])
augmented_train[target] = le.transform(augmented_train[target])
test_set[target] = le.transform(test_set[target])

# Train a Random Forest model on the original training dataset.
model_original = train_rf_model(original_train, features, target)
metrics_original = evaluate_model(model_original, test_set, features, target)

# Train a Random Forest model on the augmented training dataset.
model_augmented = train_rf_model(augmented_train, features, target)
metrics_augmented = evaluate_model(model_augmented, test_set, features, target)

unique_labels = np.sort(metrics_original['y_test'].unique())
# Now, inverse transform these labels using the same LabelEncoder 'le' used earlier
target_names = [str(x) for x in le.inverse_transform(unique_labels)]

# Compare the performance of the two models.
compare_models(metrics_original, metrics_augmented, target_names)
"""

"""
import pandas as pd
from MachineLearningModels.MultiClassKNN import train_knn_model, evaluate_model, compare_models
from sklearn.preprocessing import LabelEncoder
import numpy as np

# Load CSV files for the Yeast dataset.
original_train = pd.read_csv("OutputTrainingSets/original_trainVAEForest.csv")
augmented_train = pd.read_csv("OutputTrainingSets/augmented_trainVAEForest.csv")
test_set = pd.read_csv("OutputTrainingSets/test_setVAEForest.csv")

# Define the feature columns and target.
features = numeric_features
categorical_features = ["Cover_Type"]

# Re-map the target column so that labels are contiguous.
le = LabelEncoder()
original_train[target] = le.fit_transform(original_train[target])
augmented_train[target] = le.transform(augmented_train[target])
test_set[target] = le.transform(test_set[target])

# Train a KNN model on the original training dataset.
model_original = train_knn_model(original_train, features, target)
metrics_original = evaluate_model(model_original, test_set, features, target)

# Train a KNN model on the augmented training dataset.
model_augmented = train_knn_model(augmented_train, features, target)
metrics_augmented = evaluate_model(model_augmented, test_set, features, target)

unique_labels = np.sort(metrics_original['y_test'].unique())
# Now, inverse transform these labels using the same LabelEncoder 'le' used earlier
target_names = [str(x) for x in le.inverse_transform(unique_labels)]

# Compare the performance of the two models.
compare_models(metrics_original, metrics_augmented, target_names)
"""